By: Prof Moses Azong Cho
I would like to share a template that presents how we ought to perform accuracy assessment for land cover studies. We may note that the area of land cover for each class obtained directly from a classified map may be very different from the true area of the class. As noted in the paper of Olofsson et al. (2013), an error-adjusted estimator of area can be easily produced once an accuracy assessment has been performed and an error matrix constructed.
The procedure presented in the table leads to the establishment the unbiased estimator of the total area of each class i.e. the error adjusted estimator of the area because it includes the area of the map omission error of each class and leaves out the commission error. Furthermore, +/- 95% confidence interval are generated for each class.
It is recommended that the producer, user and overall accuracies are computed from the error matrix of estimated area proportions and not from the traditional confusion matrix as the latter does not often take the proportion of each class in the study area into account in the accuracy assessment.
Please adjust the template as appropriate – according to your study
Change detection matrices should be constructed in a similar manner – see Olofsson et al. 2013 for a good example. This should avoid wrong conclusions about the actual area of change.
As one of the associate editors of the International Journal of Applied Earth observation and Geoinformation, I cannot accept a paper for publication if proper accuracy assessment as described above is not conducted.
The procedure as described by Olofsson et al. 2013 is the recommended standard by Intergovernmental Panel on Climate Change (IPCC).
Share the template with colleagues, student and land use land cover change practitioners. Click here to download
